5.7 KiB

Raw Blame History

selderee

Selectors decision tree - pick matching selectors, fast.

What is it for

The problem statement: there are multiple CSS selectors with attached handlers, and a HTML DOM to process. For each HTML Element a matching handler has to be found and applied.

The naive approach is to walk through the DOM and test each and every selector against each Element. This means O(n*m) complexity.

It is pretty clear though that if we have selectors that share something in common then we can reduce the number of checks.

The main selderee package offers the selectors tree structure. Runnable decision functions for specific DOM implementations are built via plugins.

Limitations

Pseudo-classes and pseudo-elements are not supported by the underlying library parseley (yet?);
General siblings (~), descendants ( ) and same column combinators (||) are also not supported.

`selderee` vs `css-select`

css-select - a CSS selector compiler & engine.

Feature	`selderee`	`css-select`
Support for `htmlparser2` DOM AST	plugin	+
"Compiles" into a function	+	+
Pick selector(s) for a given Element	+
Query Element(s) for a given selector		+

Packages

Package	Version	Folder	Changelog
selderee		/packages/selderee	changelog
@selderee/plugin-htmlparser2		/packages/plugin-htmlparser2	changelog

Install

> npm i selderee @selderee/plugin-htmlparser2

Documentation

Usage example

const htmlparser2 = require('htmlparser2');
const util = require('util');

const { DecisionTree, Treeify } = require('selderee');
const { hp2Builder } = require('@selderee/plugin-htmlparser2');

const selectorValuePairs = [
  ['p', 'A'],
  ['p.foo[bar]', 'B'],
  ['p[class~=foo]', 'C'],
  ['div.foo', 'D'],
  ['div > p.foo', 'E'],
  ['div > p', 'F'],
  ['#baz', 'G']
];

// Make a tree structure from all given selectors.
const selectorsDecisionTree = new DecisionTree(selectorValuePairs);

// `treeify` builder produces a string output for testing and debug purposes.
// `treeify` expects string values attached to each selector.
const prettyTree = selectorsDecisionTree.build(Treeify.treeify);
console.log(prettyTree);

const html = /*html*/`<html><body>
  <div><p class="foo qux">second</p></div>
</body></html>`;
const dom = htmlparser2.parseDocument(html);
const element = dom.children[0].children[0].children[1].children[0];

// `hp2Builder` produces a picker that can pick values
// from the selectors tree.
const picker = selectorsDecisionTree.build(hp2Builder);

// Get all matches
const allMatches = picker.pickAll(element);
console.log(util.inspect(allMatches, { breakLength: 70, depth: null }));

// or get the value from the most specific match.
const bestMatch = picker.pick1(element);
console.log(`Best matched value: ${bestMatch}`);

Example output

▽
├─◻ Tag name
│ ╟─◇ = p
│ ║ ┠─▣ Attr value: class
│ ║ ┃ ╙─◈ ~= "foo"
│ ║ ┃   ┠─◨ Attr presence: bar
│ ║ ┃   ┃ ┖─◁ #1 [0,2,1] B
│ ║ ┃   ┠─◁ #2 [0,1,1] C
│ ║ ┃   ┖─◉ Push element: >
│ ║ ┃     └─◻ Tag name
│ ║ ┃       ╙─◇ = div
│ ║ ┃         ┖─◁ #4 [0,1,2] E
│ ║ ┠─◁ #0 [0,0,1] A
│ ║ ┖─◉ Push element: >
│ ║   └─◻ Tag name
│ ║     ╙─◇ = div
│ ║       ┖─◁ #5 [0,0,2] F
│ ╙─◇ = div
│   ┖─▣ Attr value: class
│     ╙─◈ ~= "foo"
│       ┖─◁ #3 [0,1,1] D
└─▣ Attr value: id
  ╙─◈ = "baz"
    ┖─◁ #6 [1,0,0] G
[ { index: 2, value: 'C', specificity: [ 0, 1, 1 ] },
  { index: 4, value: 'E', specificity: [ 0, 1, 2 ] },
  { index: 0, value: 'A', specificity: [ 0, 0, 1 ] },
  { index: 5, value: 'F', specificity: [ 0, 0, 2 ] } ]
Best matched value: E

Some gotcha: you may notice the check for #baz has to be performed every time the decision tree is called. If it happens to be p#baz or div#baz or even .foo#baz - it would be much better to write it like this. Deeper, narrower tree means less checks on average. (in case of .foo#baz the class check might finally outweigh the tag name check and rebalance the tree.)

Development

Targeting Node.js version >=14.

Monorepo uses NPM v7 workspaces (make sure v7 is installed when used with Node.js v14.)

5.7 KiB Raw Blame History